icon

Argo Floats android application¶

Analytics stuff¶

In [1]:
import pandas as pd
import glob, folium, branca, json
import numpy as np
import matplotlib.pyplot as plt

import holoviews as hv
import panel as pn
import panel.widgets as pnw

hv.extension('bokeh')
In [2]:
from bokeh.themes.theme import Theme

theme = Theme(
    json={
'attrs' : {
    'Figure' : {
        'background_fill_color': '#535353',
        'border_fill_color': '#535353',
        'outline_line_color': '#444444',
    },
    'Grid': {
        'grid_line_dash': [6, 4],
        'grid_line_alpha': .3,
    },

    'Axis': {
        'major_label_text_color': 'white',
        'axis_label_text_color': 'white',
        'major_tick_line_color': 'white',
        'minor_tick_line_color': 'white',
        'axis_line_color': "white"
    }
  }
})
hv.renderer('bokeh').theme = theme
In [3]:
print(np.datetime64('now'))

2022-09-16T02:47:10
In [4]:
# PUT EVERY COUNTRY REPORTS IN ONE DATAFRAME
files = glob.glob('data/install*country.csv')
files.sort()
ds=pd.DataFrame()
for f in files:
    ds = pd.concat([ds,pd.read_csv(f,encoding = 'utf-16')])
ds = ds.reset_index()    
ds = ds.drop(columns=['index','Daily Device Upgrades','Total User Installs','Active Device Installs','Install events','Update events','Uninstall events'])
ds['Date']=pd.DatetimeIndex(ds['Date'])
ds['WO-ui']=ds['Daily User Installs'].cumsum()
ds['WO-di']=ds['Daily Device Installs'].cumsum()
ds['WO-uu']=ds['Daily User Uninstalls'].cumsum()
ds['WO-du']=ds['Daily Device Uninstalls'].cumsum()
In [5]:
countries = np.unique(ds['Country'][~ds['Country'].isna()])
for country in np.unique(ds['Country'][~ds['Country'].isna()]):
    ds[country+'-di']=ds[ds['Country']==country]['Daily Device Installs'].cumsum()
    ds[country+'-ui']=ds[ds['Country']==country]['Daily User Installs'].cumsum()
    ds[country+'-du']=ds[ds['Country']==country]['Daily Device Uninstalls'].cumsum()
    ds[country+'-uu']=ds[ds['Country']==country]['Daily User Uninstalls'].cumsum()
ds = ds.fillna(method='ffill')
ds = ds.fillna(0)
ds.tail()
Out[5]:
Date Package Name Country Daily Device Installs Daily Device Uninstalls Daily User Installs Daily User Uninstalls WO-ui WO-di WO-uu ... UZ-du UZ-uu VE-di VE-ui VE-du VE-uu ZA-di ZA-ui ZA-du ZA-uu
11833 2022-09-12 com.kb.android.argo PA 0 0 0 0 190 212 139 ... 0.0 1.0 4.0 4.0 0.0 5.0 1.0 1.0 0.0 0.0
11834 2022-09-12 com.kb.android.argo RU 0 0 0 0 190 212 139 ... 0.0 1.0 4.0 4.0 0.0 5.0 1.0 1.0 0.0 0.0
11835 2022-09-12 com.kb.android.argo TR 0 0 0 0 190 212 139 ... 0.0 1.0 4.0 4.0 0.0 5.0 1.0 1.0 0.0 0.0
11836 2022-09-12 com.kb.android.argo US 0 0 0 0 190 212 139 ... 0.0 1.0 4.0 4.0 0.0 5.0 1.0 1.0 0.0 0.0
11837 2022-09-12 com.kb.android.argo ZA 0 0 0 0 190 212 139 ... 0.0 1.0 4.0 4.0 0.0 5.0 1.0 1.0 0.0 0.0

5 rows × 275 columns

Installations cumulation¶

In [6]:
A2 = pd.read_csv('A2codes.csv',index_col=1)
A2i = pd.read_csv('A2codes.csv',index_col=0)

labels = sorted([A2['Name'][x] for x in np.hstack([countries,'WO'])])

label = pnw.Select(name='Country', value='World', options=labels)
ptype = pnw.Select(name='Type', value='Install', options=['Install','Uninstall'])


@pn.depends(label.param.value,ptype.param.value) 
def create_figure(label,ptype):
    code = A2i['Code'][label]  
    if code != 'WO':
        dsi = ds[ds['Country']==code]
    else:
        dsi=ds
    
    hv_data = hv.Table(dsi, ['Date'])    
    p1 = hv_data.to.curve(['Date'], [code+'-u'+ptype[0].lower()],label='Users').opts(color='#2acaea')
    p2 = hv_data.to.curve(['Date'], [code+'-d'+ptype[0].lower()],label='Devices').opts(color='#d71e3e')      
    p = p1*p2
    p.opts(hv.opts.Curve(width=700, height=400,show_grid=True),
           hv.opts.Overlay(legend_position='top_left'))
    
    return p

#Panel dataframe looks better than holoview's
@pn.depends(label.param.value,ptype.param.value)
def create_table(label,ptype):
    code = A2i['Code'][label]  
    if code != 'WO':
        dsi = ds[ds['Country']==code]
    else:
        dsi=ds
    return pnw.DataFrame(dsi[['Date',code+'-u'+ptype[0].lower(),code+'-d'+ptype[0].lower()]].groupby('Date').max(),height=400, widths=180, autosize_mode='none')


widgets = pn.WidgetBox(label, ptype, width=170)
pn.Row(widgets, create_figure, create_table)
Out[6]:

World choropleth from install¶

In [7]:
import json
with open('countries.json') as f:
    gj = json.load(f)
In [8]:
df = pd.DataFrame(ds.groupby('Country').sum()['Daily User Installs'])
color_scale = branca.colormap.linear.viridis.scale(0,20)
map_dict = df.to_dict()

def get_count(ISO_A2,ADMIN):
    value = map_dict['Daily User Installs'].get(ISO_A2)
    if value is None:
        return ADMIN+" : No install" 
    else:        
        return ADMIN+" : "+str(value)

for i in range(len(gj['features'])):
    gj['features'][i]['properties']['INSTALL'] = get_count(gj['features'][i]['properties']['ISO_A2'],gj['features'][i]['properties']['ADMIN'])
    
def get_color(feature):
    value = map_dict['Daily User Installs'].get(feature['properties']['ISO_A2'])
    if value is None:
        return 'white' # MISSING -> white
    else:
        #print(feature['properties']['ADMIN']+' : '+str(value))
        return color_scale(value)

m = folium.Map(
    location = [0, 0], 
    tiles="cartodbpositron",
    zoom_start = 2
)

folium.GeoJson(
    data = gj,
    popup=folium.GeoJsonPopup(fields=['INSTALL']),
    style_function = lambda feature: {
        'fillColor': get_color(feature),
        'fillOpacity': 0.7,
        'color' : 'None',
        'weight' : 1,
    }    
).add_to(m)
m.add_child(color_scale)
m
Out[8]:
Make this Notebook Trusted to load map: File -> Trust Notebook